Old Dominion University ODU Digital Commons
OEAS Faculty Publications Ocean, Earth & Atmospheric Sciences
2010 Benthic Ecology From Space: Optics and Net Primary Production in Seagrass and Benthic Algae Across the Great Bahama Bank Heidi M. Dierssen
Richard C. Zimmerman Old Dominion University, [email protected]
Lisa A. Drake Old Dominion University
David J. Burdige Old Dominion University, [email protected]
Follow this and additional works at: https://digitalcommons.odu.edu/oeas_fac_pubs Part of the Ecology and Evolutionary Biology Commons, Marine Biology Commons, Oceanography Commons, and the Sedimentology Commons
Repository Citation Dierssen, Heidi M.; Zimmerman, Richard C.; Drake, Lisa A.; and Burdige, David J., "Benthic Ecology From Space: Optics and Net Primary Production in Seagrass and Benthic Algae Across the Great Bahama Bank" (2010). OEAS Faculty Publications. 71. https://digitalcommons.odu.edu/oeas_fac_pubs/71 Original Publication Citation Dierssen, H.M., Zimmerman, R.C., Drake, L.A., & Burdige, D. (2010). Benthic ecology from space: Optics and net primary production in seagrass and benthic algae across the Great Bahama Bank. Marine Ecology Progress Series, 411, 1-15. doi: 10.3354/ meps08665
This Article is brought to you for free and open access by the Ocean, Earth & Atmospheric Sciences at ODU Digital Commons. It has been accepted for inclusion in OEAS Faculty Publications by an authorized administrator of ODU Digital Commons. For more information, please contact [email protected]. VoL 411: 1-15, 2010 MARINE ECOLOGY PROGRESS SERIES Published Jaly 29 doi: 10.3354/meps0866S Mar Ecol Prog Ser fOPENl ~ FEATURE ARTICLE
Benthic ecology from space: optics and net primary production in seagrass and benthic algae across the Great Bahama Bank
1 2 2 3 2 Heidi M. Dierssen ·· , Richard C. Zimmerman , Lisa A. Drake • , David Burdige
1 Department of Marine Sdences/Geography, University of Connecticut, Groton, Connecticut 06340, USA 2Department of Ocean, Earth and Atmospheric Sdences, Old Dominion University, 4600 Elkhorn Avenue, Norfolk, Virginia 23529, USA 3.Present address: Science Applications tntemational Corporation, clo Naval Research Laboratory, Key West, Florida 33041, USA
ABSTRACT: Development of repeatable and quantita tive tools are necessary for determining the abundance and distribution of different types of benthic habitats, de tecting changes to these ecosystems, and determining their role in the global carbon cycle. Here we used ocean color remote sensing techniques to map different major groups of primary producers and estimate net primary productivity (NPP) across Great Bahama Bank (GBB). Field investigations on the northern portion of the GBB in 2004 revealed 3 dominant types of benthic primary pro ducer.;: seagrass, benthic macroalgae, and microalgae at tached to sediment. Laboratory measurements of NPP ranged from barely net autotrophic for grapestone sedi ment with thin microalgal biofilm to highly productive for dense accumulations of brown macroalgae. A logarith Pseudo-true color image from the MODIS Aqua sensor at mic relationship between NPP and green seafloor re 250 m resolution shows the multihued reflections from flectance described the general trend in NPP across shallow waters in the southern Bahamas. various benthic constituents. Using a radiative transfer based approach, satellite-derived estimates of NPP for Image: C. Buonass.issi and the region totaled-2 x 1013 gC yr1 across the GBB. The NASA Oceaa Biology Processing Group prevailing benthic habitat was mapped as sediment with little to no microalgal biofilm. Moderate to dense sea grass meadows of Thalassia testudinum were the domi INTRODUCTION nant primary producers and contributed over 80 % of NPP in the region. H the vast majority of seagrass leaves In 1918, Johannes Petersen published the first spa decompose in the primarily carbonate sediments, car tial estimates of seagrass production along a 7000 km2 bonate dissolution processes associated with this decom basin off the coast of Denmark and reported that it position may result in sequestration of seagrass above was more than 4 times the quantity of hay annually and below-ground carbon into the bicarbonate pool (2.4 13 produced by Danish fields and meadows (Petersen x 10 gC yr1), where it has a residence time on the order of tens of thousands of years. 1918). Despite this early effort, global assessments of productivity from seagrass and other benthic produc KEY WORDS: Primary productivity · Remote sensing · er.; remain elusive and their contributions to regional Ocean color · Inherent optical properties · Benthic and global biogeochemical cycles remain poorly reflectance · Bahamas · Grapestone sediment quantified (Smith 1981, Short & Wyllie-Echeverria --- Resale or republication not permitted without ___ 1996). written consent of the publisher
•Email: [email protected] 0 Inter-Research 2010 · www.int-res.com 2 Mar Ecol Prog Ser 411: 1-15, 2010
Benthic producers can have relevance to climate ing 'whitings' that are visible from space (Robbins et studies both in terms of the amount of productivity al. 1997, Dierssen et al. 2009a). These turbid, wind-dri and the fate of that productivity. For example, sea ven whiting events can obscure the benthos from view, grasses may cover roughly 6 x H>5 km2 of the global but cover only a relatively small area per episode. seafloor and their net primary production (NPP) is Hence, the expansive GBB is, for the most part, opti estimated to be 0.5 PgC yr1 (Duarte & Cebrian 1996, cally shallow, in that reflectance of the seafloor con Mateo et al. 2006). This reflects roughly 1 % of NPP of tributes to the rellected light measured by a satellite or the ocean's phytoplankton (Longhurst et al. 1995). aircraft. The role benthic producers play in both the local ecol As with phytoplankton in the open ocean, variations ogy and biogeochemical cycles, however, also depends in benthic environments change the ocean color on how much production is left unremineralized by expressed at the sea surface, which can be exploited to herbivores and decomposers to accumulate in the estimate the distributions of the respective seafloor seabed (Short 1980, Mateo et al. 2006). Along the constituents. Many methods have been developed for Exumas portion of the Great Bahama Banlc (GBB), conducting remote sensing of seagrasses and other highly productive mats of negatively buoyant benthic benthic producers (reviewed by Kelly 1980, Dekker et macroalgae may be transported into the deep Tongue al. 2006). Such approaches vary depending on the of the Ocean in pulsed exports exceeding 7 x 1010 g of spectral resolution of available imagery and can in carbon (Dierssen et al. 2009b). Such events are equiv volve look-up tables (Dierssen et al. 2003, Louchard et alent to the daily carbon flux of phytoplankton bio al. 2003, Mobley et al. 2005), hyperspectral optimiza mass in the pelagic tropical North Atlantic, and 0.2 to tion techniques (Lee et al. 1999), and other radiative 0.8 % of daily carbon Hux from the global ocean. transfer-based approaches (Dekker et al 2005). The Moreover, different species of seagrass may play amount and spectral shape of reflected light is influ different ecological roles based on whether the major enced by water depth, water column optical proper ity of the biomass accumulates locally within the bed ties, and rellectance signatures of the underlying or is transported out of the ecosystem (Zieman et al. benthos (Armstrong 1993, Maritorena et al. 1994). 1979, Burdige & Zimmerman 2002, Burdige et al. However, methods depend on the spectral resolution 2008). Hence improved methods for quantifying large available from a sensor (Phinn et al. 2008), in addition scale benthic ecology are warranted. to the spatial resolution, and the ability to retrieve cal Quantifying productivity of benthic ecosystems is ibrated and atmospherically corrected imagery over challenging due to the high degree of fine-scale vari the optically shallow region. ability in the seafloor features in many environments In the present study, field data were used to develop (cm to m scale) and the limited perspective available an ocean color remote sensing model to characterize from traditional in situ sampling methods. Remote the benthic habitats of the GBB. Using a radiative sensing of benthic environments from space offers the transfer~based approach, the distribution of broad cat potential for filling in this gap (reviewed in Dekker et egories of primary producer bilbitats (seagrass, algae, al. 2006, Phinn et al. 2008); however, such methods and sand) were estimated from variations in remotely require considerable in situ measurements to develop sensed ocean color, from hues of light blue (primarily approaches to differentiate among types of benthic sand) to green (vegetated), across the GBB and com producers (seagrass, algal-covered sand, sediment) bined with empirical relationships between primary and to differentiate the optical signatures of the water producer abundance and benthic productivity to esti column and the benthos. mate the contribution of these highly productive shal The GBB is an ideal experimental platform to develop low seas to the global carbon cycle. and test remote sensing approaches for estimating NPP of benthic habitats. The water is consistently shal low across the bank (3 to 10 m) and large regions of MATERIALS AND METHODS 'green' seafloor are observable from space (Fig. 1). Moreover, the water overlying much of the GBB is rel The present study was conducted on the GBB in atively oligotrophic, without significant water column March 2004 aboard the RV 'Walton Smith' (Fig. 1). phytoplankton biomass. Water column optical proper Twenty-five stations were sampled across the northern ties are primarily dominated colored dissolved organic portions of GBB near Andros and along the Exumas matter produced by benthic organisms, such as sea Archipelago (Fig. 1). At each station, divers were grass (Boss & Zaneveld 2003, Otis et al. 2004). How deployed in the water to survey the site, quantify sea ever, on the western portion of the GBB near Andros grass densities if present, and collect sediment cores. Island, fine lime mud and pellet mud sediments can be Water column optical and physical measurements and periodically suspended into the water column, produc- hyperspectral reflectance measurements were con- Dierssen et al.: Benthic productivity from space 3
79" w 11· w Optical measurements. A variety of opti cal measurements were conducted to char acterize how the water column and sea Seagrass (Dense) floor absorb, scatter, and reflect light. The Seagrass (Moderate) reader is referred to Mobley (1994) for a v Seagrass (Sparse) more detailed characterization of optical • Brown macroalgae terminology. An ac-9 Plus package (WET A Red macroalgae Labs) was deployed at each station to mea • Grapestone sediment Mud 9ediment sure light absorption and attenuation of the Sponges water column at a fixed depth of 2 to 3 m below the sea surface. Salinity and temper ature were measured using an SBE 19 (Sea-Bird Electronics) integrated into the package. The particulate and dissolved absorption and attenuation coefficients, apg and c,.g, respectively, were corrected for temperature, salinity, and scattering effects (Dierssen et al. 2009a). Drift offsets were obtained from a post-cruise clean water calibration. Total particulate scattering (hp) was calculated as the difference between Gig and a,g. assuming negligible scattering from dissolved material. The particulate backscattering coefficient ( bt,p) refers to all the photons that have been redirected in the backward direction due to scattering from particles in the water and, to a first order, is positively correlated to the total concentration of particles in the water (re viewed in Stramski et al. 2004). The ECO VSF (WET Labs) provided backscattering measurements at 3 angles (100, 125, and Fig. 1. SeaWJ.FS true color image (6 March 2004) of the tropical West Atlantic 150°) and 3 wavelengths (470, 532, and reveals extensive areas of green and blue-colored optically shallow water 650 nm). Data were corrected for light across the entire Great Bahama Banks. Dominant benthic type at each field attenuation effects using concurrent ac-9 station is identified in the key data and extrapolated from 90 to 180° using a third-order polynomial. The backscatter ducted in situ. Sediment, seagrass, and macroalgae ing ratio (bi,p) was calculated as the ratio of bi,p to hp. samples were collected by divers and analyzed further Remote sensing reflectance (R.s) represents the spec in the laboratory. Reflectance measurements were con tral distribution of light emerging from the sea surface ducted on the sediment cores collected at all stations. normalized to that incident upon the sea surface and Oxygenic photosynthesis and respiration was mea was measured at each station using a Field Spec Pront sured at a subset of the stations representing the differ VNIR-NIRl portable spectrometer system (Analytical ent habitats: turtle grass leaves of all ages present Spectral Devices). A sequence of measurements was within a shoot were analyzed with epiphytes from made with an 8° foreoptic focused at a 40 to 45° angle Stns WS12, WS13, WS16, and WS20; the macroalgae sequentially on a gray plaque, sea surface, and sky. Colpomenia sp. was analyzed from Stn WS7 where it The parameter Rn was calculated as the water-leaving was found in organized benthic windrows caused by radiance normalized to downwelling irradiance reach Langmuir circulation (Dierssen et al. 2009b); grape ing the sea surface following Mobley et al. (1999). Val stone sediments with extensive microalgal cover were ues of Rn exclude photons which have reflected off the analyzed from Stn WS6. Seagrass productivity was also sea surface and do not penetrate the water surface (i.e. estimated from in situ leaf growth rates and compared sun glint). Residual reflected sky radiance was cor to field measurements conducted in 2002 and 2003 rected using reflectance differences between 715 and near Lee Stocking Island, Bahamas (Dierssen et al. 735 nm, an approach specifically developed for coastal 2003). waters (Gould et al 2001). ' Mar Ecol Prog Ser 411: 1-15, 2010
Seagrass morphology, density, and growth. Shoot trode drift was measured in chambers filled with air density of the seagrasses Tbalassia testudinum (turtle saturated seawater prior to the addition of each sample. grass), Syringodium filiforme (manatee grass), and Respiration of each sample was measured in the dark Halodule wrightii (shoal grass) were estimated at each using air-saturated seawater. At the end of the respira station using 15 to 20 randomly located 0.04 m1 tion period.. Oz content of the incubation medium was quadrats within a 20 m radius around a central buoy, further reduced to about 50% of air saturation by bub 1 spanning roughly 1000 m at each station. One shoot of bling with a gas mixture of C01 (350 ppm) and N2 (bal each species (if present) was collected randomly from ance) before commencing photosynthesis measure each quadrat and the length (nearest mm) and width ments The light-saturated rate of photosynthesis (Pm) (nearest 0.01 mm) of each leaf was measured in the was measured by exposing each sample to -600 µmol laboratory. Leaf area index (LAI) was estimated from quanta m-1 s-1 of white light provided by a fiber optic leaf morphometrics and shoot density counts as the light source using a variable intensity incandescent 2 average one-sided leaf area per surface area (m ) of lamp (Dolan-Jenner Industries, Fiber-Lite Series 180). substrate (Zimmerman 2003). Incubation irradiance was measured using a factory Reflectance of individual turtle grass leaves were calibrated quantum irradiance meter (Biospherical In measured in the laboratory using a Shimadzu UV2101- struments, QSL-100). Sample weights were determined PC scanning spectrophotometer fitted with an inte after drying at 65°C for - to d. NPP was estimated as grating sphere. Sediment cores were collected by suming photosynthesis was light saturated for 12 h ct-1 divers and the surface reflectance was estimated in the and the photosynthetic quotient (02:COz) was 1.2, laboratory using the Field Spec-Pro with the bare fiber following the procedure outlined in Dierssen et al. optic probe positioned over the illuminated sample and (2009b). Seagrass leaves were assumed to comprise referenced against a white spectralon plaque (99.9 % 0.05 g FW cm-1 of leaf area, which was derived from reflectance) positioned similarly. For each core, 5 dif turtle grass samples collected at Lee Stocking Island ferent spectra were recorded and averaged. Because (Zimmerman 2003). The drifting macroalga Colpome of the way light refracts as it transfers through an nia sp. was determined to be 3 to 10 cm in thickness dis air-water interface, these measurements approximate tributed in ordered windrows (Dierssen et al. 2009b) the spectral shape of reflectance, but the absolute and assumed to be 2 cm in thickness averaged over spectral magnitude may vary somewhat from the non-windrow area without windrows. The macroalgae reflectance of submerged cores. was negatively buoyant, but the density was conserva Leaf growth rates were measured at 2 stations popu tively assumed to be that of seawater (1025 kg m-3). lated by turtle grass. One shoot from each quadrat was Image processing. A full resolution (1 km) Sea-view marked with a 20 gauge hypodermic needle at the top ing Wide Field-of-View (SeaWiFS) Level 1A Local of the leaf sheath (Zieman & Wetzel 1980) and tagged Area Coverage image from 6 March 2004 provided a with a length of pink flagging tape for subsequent relatively cloud-free view of the GBB. Standard atmos identification. The shoots were collected 2 wk after pheric correction routines for SeaWiFS, which were marking and the new growth of each leaf was mea developed for optically-deep open ocean water, cannot sured. NPP was estimated for each site as the product be applied accurately to the optically shallow GBB. 1 of LAI and leaf growth rate (% ct- ) and converted to a Consequently, R.. was derived at each pixel in the mass-specific rate using measured ratios for shoot image by applying an iterative scheme to first process fresh weight (FW) and dry weight (OW), which ranged the image using the SeaWiFS multiple-scattering from 0.09 to 0.22 g DW g-1 FW. Average conversions of atmospheric correction algorithm (Gordon 1997). The 0.05 g FW cm-2, 0.16 g OW g-1 FW, and 0.35 gC g-1 OW atmospheric correction scheme from neighboring deep were used (Zimmerman 2003). water was identified and the entire image was then Primary producer metabolism. Oxygenic photosyn reprocessed with SeaWiFS Data Analysis System using thesis and respiration of turtle grass leaves, brown al a fixed aerosol model for the region. This type of gae Colpomenia sp. thalli, and grapestone sediments atmospheric correction is only an approximation of the were measured polarographically in clear glass water atmospheric properties over shallow water. Errors in jacketed incubation chambers (5 ml volume). For the aerosols, whitecaps, and sun glint estimated over deep turtle grass leaves, 3 replicates of the different leaf ages water can lead to consistent biases in retrieved were conducted at Stn WS04, 2 replicates at Stns WS12 reflectances over shallow water. A linear correction and WS13, and 1 measurement for each leaf age at was derived for each wavelength where the offset Stns WS02, WS03, WS16, and WS20. Three replicates represents errors due to path radiance and the slope were conducted for Colpomenia sp. and only 1 mea represents errors in the transmission of water-leaving surement for the grapestone sediment. Temperature in radiance through the atmosphere (Gordon 1997). side the chambers was maintained at 22°C and elec- Linear regression was conducted with R.. derived from Dierssen et al.: Benthic productivity from space 5
the SeaWiFS sensor and in situ measurements from a subset of 6 stations considered to be consis "' ~~~~S~8oS8S88SS8o~S .,~-·e.. c:ic:ic:ic:ic:ic:ic:ic:ic:ic:ic:ic:ic:ic:ic:ic:ic:ic:ic:i tent at km scales from our field observations. Some !;.., ..o <.J ...... stations, for example, had patchy distributions of Ql .. 8 ~~~~~~~~~~~~~~~~~~~ VJ - o-Noooo-00-000-0-0 seagrass and were excluded from this analysis. The image was then georeferenced and warped based on a complete geometry model of the earth and satellite orbit. Estimates of bottom reflectance, - II) were modeled from the R,. (in situ values used c:i c:i Ra. I I I I it it I I I I I I I I I I I I I for model development or satellite-derived values ...;- ...;II) for model application) following the look-up table approach outlined in Dierssen et al. (2003), as dis gg cussed further in the 'Results'. c:i c:i I I I I it it I I I I I I I I I I I I I 00II) - RESULTS c:i c:i
Benthic characterization :: ~ ~~~ We encountered a variety of benthic habitats it I it I it it it I I I I I I I I I I I I .., ~ .,,_Cll across the study area, including those predomi .. c:i ai cxi c:i nantly comprised of seagrass, benthic macroalgae, - N - and sediment microalgae. Seagrass meadows (Fig. 1) were concentrated along the outer perimeter ~ "'~ c:i c:i of the bank in areas experiencing regular tidal ex it I it I I I I I I I I I I I I I change with Atlantic waters. Stations to the west of ~ C! Andros lsland were comprised of fine pellet muds, but also contained sparse distributions of seagrass. Along the Exumas, 4 different types of benthic com munities were found: seagrass, brown macroalgae, I:' N I I l:'
found further north in this area at depths up to 10 m despite the fact that different stations contained differ (Fig. 1). Along the Exumas, meadows of turtle grass ent mixtures of seagrass species. The residual variabil occurred primarily at 3 to 5 m depth in water bordering ity among stations was due to differences in shoot mor the channels near the islands with LAI ranging up to 2. phometry, including leaf length and width. For both At 2 stations close to the Exuma Islands, shoal grass seagrass species, canopy heights were significantly was found interspersed with turtle grass. Sparse disbi taller at the Andros stations compared to the Exumas. butions of turtle grass were found directly west of Shoot leaf area of turtle grass, calculated as the sum of Andros Island growing in the mud sediment (LAI< 0.2). the one-sided area of all leaves on each shoot. averaged Isolated patches of dense turtle grass occurred sporad 29 cm2 shoor1 for Andros stations and only 15 cm2 ically along Andros (WS22) surrounded by extensive shoor1 for the Exumas populations. Shoot leaf area of flats of white sand, sponges, and isolated coral heads. manatee grass was 11 cm2 shoot""1 at Andros stations Total LAI for all species of seagrass could be pre and only 3 cm2 shoor1 at the Exumas. No trend was 2 dicted by total shoot density (shoots m- ), which ex evident between the canopy height and depth of plained over 80% of the variability in the data (Fig. 2A). occurrence (r2 = 0.17, p > 0.05, data not shown). Although the regression was heavily influenced by one The ratio of gross photosynthesis to respiration for station with very high LAI, the general relationship turtle grass averaged 5.19 ± 2.74 (Table 2), resulting in between shoot density and LAI was strongly linear, NPP ranging from 0.2 ± 0.1 gC m-2 ct-1 for relatively sparse seagrass with an LAI of 0.5 up to 0.8 ± 0.3 gC s m-2 ct-1 for dense seagrass with an LAI of 2.5. NPP from in situ leaf growth rates from 2 Andros stations (WS03 c and WS04) measured over a 2 wk period were similar 4 ...J< in magnitude (Fig. 28, solid triangles) and followed the lo( relationship: NPP =0 .4 x LAI. -~ 3 'O In comparison, NPP estimated in May- June 2002 .5 and 2003 near Lee Stocking Island on the Exumas 2 (Dierssen et al. 2003) was nearly 2-fold higher (slope of c...E 0.78) than 2004 rates (Fig. 28, circles). This higher rate could be due to differences between spring and sum .! mer sampling and greater exchange of nubients on the Exumas compared to Andros. 0 Rates of NPP for benthic algae were quite variable de 0 500 IOOO I 500 2000 2500 3000 pending on the taxon. The highest metabolism and cor Density (shoots m-2) responding NPP was measured for the phaeophyte Col pomenia sp. found growing rapidly on the mid-portion of the Exumas. Dense accumulations of this brown ~- macroalga were found as unattached free-drifting mats ~ 1.5 tumbling along the bottom with currents believed to be e generated by Langmuir circulation (Dierssen et al. u 2009b). From photosynthesis estimates on individual ~ a.. thalli, we estimated NPP to be 18 ± 4.4 gC m-2 ct-1• How a.. z ever, this estimate does not incorporate sett-shading and spectral changes in light availability within the drifting '! 0.5 2004 mats (Vergara et al. 1997, Brush & Nixon 2003), which r2 = 0.98 ·:::e may serve to diminish NPP somewhat. WI y = 0.40.r g,,J In contrast, grapestone sediments containing thin o..-::'--'=...... :""'T"~~---...... -~~.....-~~---.-~~--+ algal biofilms were barely net autotrophic, with rates 2 1 0 0.5 I l.5 2 2.S of 0.005 gC m- ct- • NPP of red algae and grapestone Leaf area index (LAI) with dense biofilm was not measured directly. Benthic Fig. 2. (A) Relationship between seagrass shoot density and microalgae within sediment from the seagrass mead estimated leaf area index (LAI) ranging from 0 to 3.5. (B) Sea ows may also contribute to NPP (Macintyre et al. 1996) grass LAI and above-ground net primary production (NPP) but was not measured directly in the present study. determined in 2002- 2003 near Lee Stocking Island, Bahamas, Around Lee Stocking Island Bahamas, sediment from in situ leaf growth rates (e), in 2QO.f. at northern Bimini stations from in situ growth rates (A), and in 20().f. from labora chlorophyll a (chl a) concentrations were 3-fold higher tory metabolism measurements (41). Dashed lines represent in dense turtle grass beds (3 to 4 µg g-1 sediment) com 95 % confidence intervals for the regression pared to adjacent sediment (Stephens et al. 2003). Oierssen et al.: Benthic productivity from space 7
Table 2. Thalassia testudinwn, Colpomenia sp. and grapestone sediment. Net primary productivity (NPPJ of benthic producers 2 measured across the Great Bahama Banlt. Leaf area index (LAI) is the area (m ) of 1-sided seagrass leaf per m2 seafloor. Pg:R: ratio of gross production to respiration; OW: dry weight
Producer· Gross oxygen evolution, Pg Pg:R Areal average 2 (µmol ~mm·• g OW-1) (gOwm- 1
Seagrass Thalassia testudinwn 0.78 :1: 0.28 (n = 23) 5.19 :I: 2.74 LAI0.5 50 0.17 :I: 0.11 LAI 1.0 100 0.35 :I: 0.13 LAI 1.5 150 0.52 :I: 0.19 LAI2.0 200 0.69 :t0.25 Macroalgae Colpomenia sp. 2.20 :1: 0.24 (n = 3) 4.71 :t0.59 2050 18 :I: 4.4• Grapestone algae 0.056(n=1) 2.12 205 0.005 •Assuming no self-shading or changes in spectral composition of light within the 2 an thick drifting macroalgal mats
Optical properties pended in the water column (Boss & Zaneveld 2003). Water from the Tongue had considerably lower absorp In order to interpret a remote sensing signal derived tion and attenuation coefficients compared to the bank. from the benthic primary producers, water column opti Andros also had higher c,,g than either the Exumas or the cal properties were measured in the surface waters at Tongue, likely due to the presence of non-absorbing all benthic stations, as well as in surface waters of the 'white' calcite fine sediment particles suspended in the Tongue of the Ocean (>1000 m). Consistent with past water column (Shinn et al. 1989, Dierssen et al. 2009a). studies and visual observations by the divers, the absorp We assumed that apg was primarily due to dissolved ab tion (Clpg) and attenuation (Cpg) coefficients were gener sorption (clg) and subtracted this from the Cpg values to ally low across the entire region, indicating high water obtain an estimate of light attenuation of particulates clarity (Fig. 3A,8). The spectral shape of the absorption alone, ep. The hyperbolic wavelength spectra of Cp for coefficient was consistent with the presence of colored Andros, Exumas, and the Tongue, respectively, were dissolved organic matter and few phytoplankton sus- 0.95 :t 0.21, 0.98 :t 0.22, and 0.8 (1 measurement only).
0.8 +-___..._ ___ ..._ __ --+ v Andros A B • Exumas O.lS 0.6 ...... o Toogue of Ocean "i 'e E 0.1 ~-- .<-- 0.4 --I! --I ~ o.os 0.2
0 0.02 +----.....____ _._ ___ _ 0.08 c D Fig. 3. Mean inherent O.OIS optical properties (:1: 1 0.06 SDJ measured at sta .2 tions from Andros, EIU 'e mas and the Tongue of :::: 0.01 } 0.04 the Ocean: (A) absorp tion by particulate and 1 dissolved material (a,.g); 0.005 O.o2 (BJ attenuation by par ticulate and dissolved material (<:pg); (Q par O+------.--e1-----.-~i:t--+ 0 ticulate backscattering 400 soo 600 700 400 soo 600 700 (bi,p); and (DJ basJtscat- tering ratio (bi.pl Wavelength (mn) 8 Mar Ecol Prog Ser 411: 1-15, 2010
The particulate backscattering data revealed a siJni des (i.e. suspended sediment) rather than biogenic lar pattern. Levels of bi.., from stations west of Andros particles (i.e. phytoplankton). Levels of bai, reached as were twice as high as stations near the Exumas high as 0.05 during an intense resuspension events or (Fig. 3C), indicative of more suspended sediment on 'whiting' which occurred at WS19 west of Andros the western arm of the GBB. This difference was likely Island (Dierssen et al. 2009a). The estimated bulk due to the larger sediment grains found along the Exu index of retraction calculated according to Twardow mas which were less prone to resuspension (Hu 2007, ski et al. (2001) on the GBB was -1.16, consistent with Dierssen et al. 2009a). Water from the Tongue was so the presence of inorganic minerogenic sediment parti devoid of particles that measurements of bi.., were cles and few organic phytoplankton. below the limit of detection for the instrument (data not Sea spectral reflectance used for remote sensing stud shown). The ratio of backscattering to total scattering ies (R... i.e. water-leaving radiance normalized to inci (bai,) was high and consistent between the Andros and dent downwelling irradiance) provided an indicator of Exumas regions (Fig. 30). Measurements averaged the water color across the visible spectrum at each sta -0.028 across all stations and no significant wave tion. Levels of R.. from selected locations at S m depth length dependence was observed. Such high values revealed quite different spectral magnitudes and shapes are consistent with the presence of minerogenic parti- depending on the seafloor composition (Fig. 4A). Grape-
0.03 ~-----~~--~---.- A l.S B e 0.02 ~ "i- -0 ~ -;- ~ r: e 0.01 z0 0.5
0 0
2 0.015 ...... c D SOS-SIOnm i Dense seagrass .._.)( 1- 0.01 J fl) .i~ 0 -e ~ > ·5 0.005 -0 - 1 '5• -2 500 600 700 0.5 E -&-Orpnic: smd. S.2 m Fig. 4. (AJ Spectral remote sensing reflectance Ill -+-Andros mud, S.4 m ~ (R,J measured at 8 stations with -5 m water 8 0.4 --GnpcslOne lilhl. S.6 m depth showing the decrease in magnitude c - - Spsse seapss. S.S m from bright sand to dark green grapestone i 0.3 - Red lll8CJQa)pe. S.6 m sediment (sr = steradianJ. (BJ The same spec c - Brown lllllmlalpc. S.3 111 trum from (AJ normalized to R,..(555J showing --Dense SClllJ'8SS. S.3 m the spectral-shift from blue- to green-peaked ~ 02 with more productive benthos. (C) Derivative 8 - Gnpcstonc d8rt.. S.7 m c spectrum of R,.. showing wavebands with ti 0.1 differences in spectral slope. (DJ Remote sens en ing reflectance over dense seagrass beds ranging in depth from 5 to 10 m. (E) Labora tory measurements of bentbic reflectance 500 600 700 IRal fcx different types of sediment, seagrass Wavelength (run) leaves,andmacroalgalb~des Dieissen et al.: Bentbic productivity from space 9
stone sediment with a green biofilm bad the lowest R.. sediment from the Exumas was twice as reflective and peaked in the green at -570 run. Stations com across the visible spectrum. The fine mud sediment prised of red and brown macroalgae had slightly from the stations around Andros was even more reflec higher R.. but similar spectral shapes. Dense seagrass tive, peaking at around 0.4 between 550 and 650 nm. beds had a similar magnitude in the green (500 to Such large differences challenge our ability to estimate 510 run), but appeared to reflect more light in the blue benthic properties from multispectral imagery, if bathy portion of the spectrum (400 to 500 run) than the other metry is unquantified. benthic producers. Remotely sensed reflectance from organic sand and Andros mud were both -5 times greater than R.. from the productive benthic habitats, Quantifying basin-wide productivity including grapestone sediment with microalgal bio film. Moreover, the spectra peaked at 500 nm com Our basic approach for modeling NPP involved esti pared to the S50 to S10 run peak observed for the other mating Rs from the imagery and then quantifying pro spectra. ductivity. To account for absorption and scattering When the spectra were normalized to R.. at 555 nm, within the water column, Rs was estimated from R.. shape differences between the various benthic types using a depth-specific look-up table derived from were readily observed (Fig. 48). At S m depth, the radiative transfer modeling (Hydrolight, Sequoia Sci reflectance peaks of high productivity stations were entific; Mobley 1994) with the measured optical prop green-shifted relative to low productivity stations. erties of the water column (Dierssen et al 2003). Aver Fourth derivatives of the spectra (Fig. 4C) revealed age water column optical properties measured during spectral regions that could potentially be used to dif the field campaign (Fig. 3, Exumas) were used in the ferentiate different benthic types with high signal-to model and assumed to be generally representative of noise imaging spectroscopy. Subtleties in spectral the GBB. Fig. SA presents the blue/green sea surface shape were most apparent in the green region of the reflectance ratio CR..[490):Rn[SSS)), used to estimate spectrum with differences observed at 505 to S10, S1S, chlorophyll in the optically deep majority of the world's S2S, and 550 nm. oceans, for stations comprised of different benthic con Highly productive dense seagrass meadows were stituents and at different depths. Many of the benthic found growing at depths >10 mat the north end of stations had lower reflective ratios indicative of chloro Andros (WS03 and WS04) and the reflectance signa phyll-containing compounds that absorb blue light. ture was dominated by blue light rather than green as However, reflectance ratios of the deeper seagrass sta the influence of water column on R.. increased with tions (9 to 10 m depth) mapped into the blue (less depth (Fig. 40). From space, therefore, the deep sea productive) region because of the impact of the over grass beds appeared more blue than green. In addi lying water column. When the spectral influence of the tion, reflectance in the red portion of the spectrum (600 overlying water column was removed by applying to 700 nm) was also 2- to 3-fold greater at the shallower the look-up table approach, seagrass, macroalgae, and stations, because more of the red light reflected from grapestone sediment with dense biofilm all grouped seafloor reaches the sea surface (Dierssen et al. 2003). together with low green benthic reflectance Rs(SSS), The reflectance of the benthos (R8 ) associated with indicative of high light absorption, while less produc each station was a combination of differently colored tive stations had higher modeled Rs(5S5) (Fig. S8). sediments and benthic constituents. Reflectance from A logarithmic relationship was developed from the seagrass leaves and macroalgae blades measured in in situ data to approximate NPP over 2 orders of mag the laboratory (Fig. 4E) had low reflectance across nitude from modeled Rs(555) (Fig. SC, R2 = 0.49, p < most of the visible spectrum. Seagrass leaves were 0.01, n = 26): log(NPP) =-7.S8R8 (SS5) + 0.65. more reflective in the blue and green compared to The relationship between benthic reflectance and brown macroalgae, which peaked further in the red at NPP separated into 3 general groups: high, medium, 600 nm. However, the seagrass leaf spectrum does not and low NPP. High NPP was associated with a light represent the reflectance from the seagrass canopy, absorbing bottom (i.e. dark reflectance); low to non which is a spectral mixture of darker seagrass leaves detectable NPP was associated with low absorbing; and brighter sediment that varies depending on the and bright bottom reflectances (>3%) were character density of the canopy and the brightness of the under istic of unvegetated sands. The NPP for brown macro lying sand (Zimmerman 2003). Sediment R.. from dif algae Colpomenia sp. was considerably higher than ferent parts of the G88 had a large range in bottom other producers, due in part to its considerable density reflectance. Measured Rs from grapestone covered on the seafloor (Dierssen et al. 2009b). NPP for red with dense algal biofilm was low in magnitude indicat algae was assumed to be similar to brown algae, but ing considerable pigment concentrations. In contrast, the density of red algae was estimated by the divers to 10 Mar Ecol Prog Ser 411: 1-15, 2010
depth-dependent look-up table. Sound 12 12 A B ings across the bank were digitized IO (Maptech 2004) and gridded to the -. IO -. 'Y• E 'Y • E same pixel size as the SeaWiFS imagery g g x (Fig. 6A). Deeper water (>8 m) occurred '5-- .JC '5-- a. v to the north of Andros and on unsampled g 6 6 ~ _.,...o._,ov regions on the southwest of the GBB. " *~i>o Shallow regions occurred to the west of 4 4 ' • • Andros Island and along much of the 2 2 Exumas (<5 m). The bathymetry was 0.6 0.8 I 1.2 1.4 0.1 0.2 0.3 corrected for tidal height to match bathy
R15(490) =R15(555) Modeled Re(SSS) metry measured at the sampling stations. Accurate application of the NPP relation 102 ship also necessitated appropriate atmo c spheric correction of the imagery. As ~ 10 1 • 'Y Seagrass (dense) ~ shown in Fig. 6B, retrieved estimates of E Seagrass (moderate) R,.. were similar in both magnitude and u 10° •v Seagrass (sparse) CG spectral shape to measurements made Brown macroalgae Cl.-- 10-• • at stations sampled during the field ;z:Cl. Red macroalgae campaign. Values of SeaWiFS-derived 10-2 *x Grapestone sediment R,..(555) were within 6 % of the measured 0 Sand and mud spectrum of the dark target and 18 % of the bright target. 0.1 0.2 0.3 Satellite-derived NPP across the GBB Modeled Re(SSS) (Fig. 6C) was low in regions west of An Fig. 5. (A) Ratio of blue to green reflectance was higher for deeper seagrass dros Island and high along its northern beds compared to similar density beds at shallower depths. (8) Modeled green border. Modeled NPP was in the range of bottom refiectance, Ra(555), from the in situ data using a radiative transfer field measurements from stations through based look-up table. (C) An inverse relationship was found between the mod eled Ra(555) and logarithmic net primary productivity (NPP) differentiating out the bank and generally followed a 1:1 regions with high, medium, and low NPP. Dashed lines represent 95% confi- line (r2 = 0.64, p < 0.01, slope= 1.2, inter dence intervals for the regression cept= 0.02; Fig. 60). The 95% confidence intervals fell within a factor of 1.7 to 2.8 over the range of NPP measured. The in be at least 1 order of magnitude lower than Colpome situ field stations were sampled at a spatial scale of nia sp. For the stations with grapestone sediment cov -1000 m2 (assuming divers sampled the seafloor ran ered with dense biofilm, NPP rates were estimated to domly within a 20 m radius around a central buoy) com be 5 times the NPP measured for the thin biofilm pared to the 1 x 1<>6 m2 area of an image pixel, and the grapestone. Past studies in this region have shown validation included only those stations with measurable chi a concentrations to be roughly 5 times greater in NPP (i.e. not bare sediment) which represented large dense biofilm grapestone compared to thin biofilm scale benthic features and had roughly similar bathym grapestone (10.2 versus 1.8 µg g-1; Stephens et al. etry across the satellite pixel compared to the measured 2003). The present study was limited to NPP measure depths. The median NPP across the GBB was 0.38 gC 2 1 ments of micro- and macroalgae at only a few stations m- A IO Bathy- c O log(NPP) 5 :r -I (gCm-2 d- 1) 0 -2 0.04+-_....__..__.__~~~~--~ IO ~ r2 =0.64, p < 0.001, ,~_-.-~-ie-,~-iFS---.1 B <'f n= 12 0.03 • E j" u .._,Cll) !. 0.02 Q., e Q., Ct: z 0.1 0.01 ] :a o+---.---..---.-:::::;=;;;;;;;;.-ii...... -~ ~ 0.01 400 450 500 550 600 650 700 750 0.01 0.1 I IO Wavelength (run) Modeled NPP (gC m-2 d-') Fig. 6. (A) Digitized and gridded bathymetry across the Great Bahama Bank from 0 to 10 m. (8) Satellite-derived remote sensing reflectance (R,J from SeaWiFS imagery l•l in comparison to field measurements from 3 diverse stations (solid lines). sr = stera 2 1 dians (C) Modeled net primary productivity (NPP, gC m· yr ) derived from SeaWiFS imagery over the entire Bahamas Banks. (D) Logarithmic comparison of satellite-modeled NPP from stations throughout the banks and estimates made during the field campaign (r = 0.64, p < 0.01). Thin dashed lines represent 95% confidence intervals and the thick dashed line is the 1:1 line 1 ows. For those regions with NPP > 0.3 gC m·2 d"" , we sects; in all cases, the sum of the diagonals is used to evaluated the retrieved Rs(440):R8 (555) ratio derived calculate the overall accuracy. Bentbic classes at 17 of from the imagery. A higher ratio was presumed to be the 23 field stations (73%) were correctly identified by predominantly seagrass and a lower ratio was either the classification technique. The off diagonals show macro- or microalgae. An analysis was conducted to errors where a portion of the map is omitted from the determine the empirical threshold of the blue:green Rs correct category (omission error) and placed in an ratio which best discriminated seagrass from the algae. incorrect category (commission error). For the seagrass The threshold was allowed to vary in 0.01 increments stations, 71 % were correctly identified and 29 o/o were from 0 to 2 until the highest agreement was reached misidentified as bare sediment. Similarly, 83 % of the between the model and the in situ identification of the sediment pixels were correctly identified and only benthos. Seagrass was modeled as those pixels with a 17% were misidentified as seagrass. No commission blue:green Rs ratio >0.7 and algae below this empirical errors occurred for algae. For algae, 50% of the 4 sta threshold (Fig. 7A). Sand was considered to be pixels tions were misidentified as either sediment (25%) or 2 1 with modeled NPP <0.3 gC m· d- • seagrass (25%). However, the technique correctly A classification accuracy matrix (Fig. 7B) illustrates identified algal stations along the central Exumas, the the model which correctly retrieved the highest num largest algal-dominated region that was sampled. ber of stations. The rows in the matrix represent the Applying this model throughout the satellite image dominant bentbic constituent identified at the field sta (Fig. 7A), seagrass accounted for 4.1x1010 gC d""1 and tions and the columns represent the bentbic class iden covered roughly 40000 km.2, while algae accounted for 1 2 tified from the satellite imagery. H all of the stations 3.3 x 1G9 gC d"" and covered only 4500 km. • A major 2 were mapped correctly, then numbers will only appear ity of the GBB (50000 km. ) was comprised of bare in the diagonal boxes where each bentbic class inter- sediment. 12 Mar Ecol Prog Ser 411: 1-15, 2010 Algae A leaf-punch studies were in reasonable agreement with the laboratory measurements of leaf metabolism and Sea grass varied by <20% for different ranges of LAI. The higher ~ Sediment rates of NPP measured in 2002-2003 near Lee Stock ing Island (see Fig. 28) were comparable to measure ments by Koch & Madden (2001) of 1.65 to 2.29 gC m-2 d-1 in a lagoon on Grand Bahama Island (little Bahama Bank) using in situ benthic chambers. Benthic algae from the GBB included brown and red macroalgae, as well as microalgae associated with sediment. Benthic diatoms have been identified as the dominant microalgae associated with all sediment types including peloidal, oolitic, and grapestone sedi ments (Stephens et al. 2003). Measured rates of algal NPP varied widely depending on the type and concen tration of algae present on the seafloor. NPP of brown Modeled B macroalgae found in thick 3 to 10 cm drifts in the Exu Classification ~" ,~ mas regions (Dierssen et al. 2009b) was very high, Accuracy Matrix !ti~ J'; ~~ upwards of 18 to 30 gC m-2 d-1 depending on macroal c.," ~ c.,Vl ] gal density. In contrast, NPP of grapestone with rela Seagrass 5 0 2 tively thin biofilm was barely net autotrophic (0.005 gC Algae 1 2 1 2 1 5 m- ct- ). Average NPP for tropical macrophyte-domi ::E Sediment 2 0 II nated ecosystems has been reported between these 2 extremes at -37 mol C m-2 yr1 or roughly 1.2 gC m-2 Fig. 7. (A) Mapped distributions of algae (either macro- or cr-1 (Gattuso et al. 1998). microalgae), seagrass, and bare sediment on the Great Within the confines of this multi-spectral analysis, no Bahama Banlt. (8) Classification accuracy matrix: diagonal values (bold) indicate the number of stations that were distinct spectral signature was found to effectively dif accurately classified in each category. Values outside the ferentiate micro- and macroalgae. Derivative analysis diagonal were misclassified stations either due to errors in the showed several spectral wavebands, particularly algorithm or variability in the spatial scale between single between 505 and 550 nm, where various producers station measurements and estimates derived over the 1 km2 pixel area could be differentiated based on spectral slope changes that may be useful should higher spectral resolution imagery become routinely available. With DISCUSSION additional spectral information (Phinn et al. 2008) and other coincident measurements and analysis (e.g. Field measurements from the GBB revealed several LIDAR, texture analysis, etc.), it may be possible to fur different types of benthic habitats: seagrass, brown ther refine the remote sensing tools to better estimate and red macroalgae, and sediment with and without the vastly different productivity rates of micro- and microalgal biofilm. Sediment with little microalgal macroalgae. This analysis suggests that highly produc biofilm or seagrass was considered the prevalent ben tive algae cover roughly 4500 km2 of the seafloor and thic habitat, covering nearly 50 000 km2 of the bank. represent 1 % of the total NPP, but this number may be Seagrass meadows were found to be the dominant pro on the low side when considering ephemeral blooms of ducer on the bank. Moderate to dense stands, primar algae and with inclusion of sediment algae within the ily of turtle grass, were mapped over 37 000 km2 of the seagrass meadows. bank and occurred primarily along the perimeters Total NPP estimated across the entire GBB was 5 x 10 1 13 of the GBB and throughout the southern bank. Due 10 gC d- or 1.8 x 10 gC yr1• Propagated error is to high winds and logistical factors, we were unable difficult to constrain given the level of complexity in to directly sample the southern portions of the bank the input data and model formulation. However, where much of the productive seagrass meadows were results from Fig. 6D suggest that the NPP estimate may identified from space; future field campaigns to this be accurate to roughly within a factor of 2. Seagrass remote centraVsouthem section of the bank will be ecosystems across the GBB represent around 0.04 % required to validate our map. of the total ocean productivity of 48 x 1015 gC yr1 Using the lower bound estimate of growth, NPP esti (Longhurst et al. 1995). U these rates are consistent mated for the densest seagrass meadows was on the across the -800 000 km2 of shallow water carbonate order of 0.8 gC m-2 cr-1• NPP estimates based on in situ sediment banks scattered throughout the world (Milli- Dierssen et al.: Benthic productivity from space 13 man 1993), productivity in carbonate systems equals from leaf growth rates represent the lower bound for roughly 0.3 % of global annual ocean productivity. This seagrass production, because they do not include the is consistent with previous estimates of global seagrass growth of below-ground roots and rhizomes or pro NPP (Mateo et al. 2006). Applying these techniques duction by other autotrophs associated with seagrass over more coastal habitats could lead to better con meadows. For these types of seagrass meadows, strained estimates of seagrass to global NPP. below-ground productivity may be as much as 35 % of Errors in the NPP estimates derived from space above-ground production (Duarte & Chiscano 1999). based imagery arise from many sources including NPP in seagrass systems is further enhanced by the atmospheric correction, bathymetry, water column larger autotrophic community consisting of phyto optical properties, and variability in estimating NPP plankton, epiphytic algae, and benthic macro- and from the different producers. The atmospherically cor microalgae. ln these clear waters, mean phytoplankton rected sea spectral reflectance derived over the bank biomass estimated from the 2004 field season was 3 was representative in magnitude and spectral shape 0.12 mg chl a m- , which roughly translates to primary across a variety of dark and bright stations sampled production of 9 gC m-2 yr1 (Behrenfeld & Falkowski field (see Fig. 68). For this analysis, bathymetry was 1997), an insignificant contribution. Seagrass system digitized from charts because good agreement could wide productivity may also be altered by epiphytic not be reached between in situ measurements and growth on the seagrass leaves and microalgae within optical methods. Bathymetry maps of the GBB from the sediment. Epiphytes do not act merely as neutral SeaWiFS (Stumpf et al. 2003) and Medium Resolu density filters situated above the seagrass leaves, but tion Imaging Spectrometer (Lee et al. 2007) imagery compete for photons with the underlying leaves (Drake showed many of the same features reproduced in the et al. 2003, Fyfe 2003). Turtle grass epiphytes from present study and these techniques hold great promise near Lee Stocking Island, Bahamas, absorbed a maxi for future retrievals. In addition, bathymetry and mum of 36 % of incident light in peak chlorophyll benthic properties can be simultaneously retrieved absorption bands and negatively impacted spectral from airborne and spacebome hyperspectral sensors photosynthesis of their seagrass hosts. For the oxygen (Kutser et al. 2006, Lesser & Mobley 2007). evolution measurements in this analysis, however, epi A look-up table based on a single set of mean water phytes were left intact and NPP estimates should be column inherent optical properties was used to model inclusive of both leaf and epiphyte productivity rates. bottom reflectance across the GBB. Estimates of Benthic microalgae from the upper layer of sediment bottom reflectance would be -15 % lower and NPP are highly variable across the Bahamas, as noted pre approximately 15 to 20% higher if the higher attenua viously, and some productivity is expected to occur tion values were used across the entire bank. The within sediment from the seagrass meadows them higher attenuation rates west of Andros, however, selves. Consequently, system-wide production within were primarily associated with fine mud sediments seagrass may be over 2-fold greater than the above that have little to no productivity and may not be rep ground seagrass production reported here. resentative of the water column overlying benthos with Finally, seagrass photosynthesis in seawater is inher higher NPP. Dense seagrass meadows can reduce cur ently COrlimited and seagrass productivity may rents inside the beds (Scoffin 1970) and can decrease increase in response to rising c~ levels in the atmos the particle load in the water (Buonassissi 2009). More phere, increasing both the Dux of organic carbon to over, the inherent optical properties used in the radia the sediments and the rates of carbonate dissolution tive transfer modeling were in the range of those mea (Zimmerman et al. 1997, lnvers et al. 2001, Palacios & sured around Lee Stocking Island (Boss & Zaneveld Zimmerman 2007). Rising atmospheric c~ may initi 2003) and may be generally representative of condi ate a positive feedback, further enhancing the seques tions in these oligotrophic waters, although some vari tration of C02 into bicarbonate by promoting dissolu ability in absorption by colored dissolved matter may tion of carbonate sediments on shallow banks that are be evident at different tidal stages (Otis et al. 2004). A heavily vegetated by seagrasses (Burdige et al. 2008). further nuanced approach to the modeling would If the vast majority of seagrass biomass decomposes in incorporate different water column optical properties the sediments, seagrass productivity across the GBB expected for different benthic substrates (Boss & Zane alone may sequester as much as 2.4 x 1013 gC yr1 (sum veld 2003) and environmental conditions (e.g. tides of above- and below-ground productivity), represent and winds) (Monahan 1993). ing nearly 1 o/o of the annual oceanic C~ uptake of In addition to the error inherent to the empirical 2.2 x 1015 gC yr1 (Feely et al. 2001). If these rates are regression, estimates of NPP may also be in error consistent across all carbonate sediment banks 2 because they do not account for total system-wide NPP (-800 000 km ; Milliman 1993), seagrass productivity within seagrass meadows. Seagrass NPP estimated may be responsible for as much as 8% of the ocean's 14 Mar Ecol Prog Ser 411: 1-15, 2010 annual net C~ uptake. Transient macroalgal blooms, grass change detection in a sballow coastal tidal Aus on the other hand, may be associated with wind-driven tralian lake. Remote Sens Environ 97:415-433 Dekker A. Brando V, Anstee J, Fyfe S, Malthus T, Karpouzli E processes along the Exumas (Dierssen et al. 2009b), (2006) Remote sensing of seagrass ecosystems: use of and their abundance has been associated with nutrient spacebome and airborne sensors. In: Larlrum AWD, Orth enrichment in coastal estuaries (Valiela et al. 1997). RJ, Duarte CM (eds) Seagrasses: biology, ecology, and Benthic microalgal communities serve to provide food conservation. Springer, Dordrecbt, p 347-359 to higher trophic levels and modify nutrient exchange Dierssen HM, Zimmerman RC, Leather RA. Downes TV, Davis CO (2003) Ocean color remote sensing of seagrass between the water column and sediments (Macintyre and bathymeby in the Bahamas Banks by high resolution et al. 1996, Berelson et al. 1998). Being able to monitor airborne imagery. Llmnol Oceanogr 48:456-463 the present and future distributions and productivity of >- Dierssen HM, Zimmerman RC, Burdige D (2009a) Optics and these various benthic producers will be important as remote sensing of Bahamian carbonate sediment whitings and potential relationship to wind-driven Langmuir Cir these regions are influenced by both natural and culation. Biogeosdences 6:487-500 anthropogenic changes. Enhanced spatial and particu Dierssen, HM, Zimmerman RC, Burdige D, Drake L (2009b) larly spectral resolution imagery from satellites may Potential export of unattached bentbic macroalgae to the afford better capabilities for such future monitoring deep sea through wind-driven Langmuir circulation. Geo and analysis. phys Res Lett 36:L04602 doi:101029/2008GL036188 Drake LA, Dobbs FC, Zimmerman RC (2003) Effects of epiphyte load on optical properties and photosynthetic Aclcnowledgemenls. Thanks are extended to S. Bailey and M. potential of the seagrasses Thalassia testudinum Banks ex Reubens from SeaDAS for help in atmospheric correction of Konig and Zostera marina L. Llmnol Oceanogr 48:456-463 the imagery, L Bodensteiner for the underwater photogra >- Duarte CM, Cebrian J (1996) The fate of marine autotrophic phy, E. Powell for assistance digitizing bathymeby, the crew production. Llmnol Oceanogr41:1758-1766 and researchers aboard the RV 'Walton Smith', and the >- Duarte CM, Chiscano CL (1999) Seagrass biomass and pro Exwna Land and Sea Park. This work was supported by duction: a reassessment. Aquat Bot 65:159-174 National Aeronautics and Space Administration's Ocean Biol Feely RA, Sabine CL, Takahashi R, Wanninkhof R (2001) ogy and Biogeocbemistry research program (NNG04GN61G Uptake and storage of carbon dioxide in the ocean: the to H.M.D. and R.C.Z.), New Energy Development Organiza global COi survey. Oceanography 14:18-32 tion of Japan. National Science Foundation (Chemical Oceano Fyfe SK (2003) Spatial and temporal variation in spectral graphy Program to D.J.B. and R.C.Z.), Caribbean Marine reflectance: Are seagrass species spectrally distinct? Um Research Center, and National Undersea Research Program. nol Oceanogr 48:464-479 Productivity measurements from Lee Stocking Island were >- Gattuso JP, Frankignoulle M, Wollast R (1998) Carbon and supported by the Office of Naval Research (N0014-97-10032 carbonate metabolism in coastal aquatic ecosystems. Annu toR.C.Z.). Rev Ecol Syst 29:405-434 >- Gordon HR (1997) Atmospheric correction of ocean color imagery in the Earth Observing System era. J Geophys LITERATURE CITED Res 102:17,081-17, 106 Gould RW, Arnone RA, Sydor M (2001) Absorption, scatter >- Armstrong RA (1993) Remote sensing of submerged vegeta ing, and remote-sensing reflectance relationships in tion canopies for biomass estimation. Int J Remote Sens coastal waters: testing a new inversion algorithm. J Coast 14:621-627 Res 17:329-341 >- Bebrenfeld MJ, Falkowski PG (1997) Photosynthetic rates Hu X (2007) Seagrass-mediated carbonate dissolution and derived from satellite-based cblorophyll concentration. early diagenesis in Bahamas Bank sediments. PhD disser Llmnol Oceanogr 42:1-20 tation, Old Dominion University, Norfolk. Virginia >- Berelson WM, Heggie D, Longmore A, Kilgore T, Nicholson >- lnvers 0, Zimmerman R, Alberte R, Perez M, Romero J (2001) G, Skyring G (1998) Bentbic nutrient recycling in Port Inorganic carbon sources for seagrass photosynthesis: an Phillip Bay, Australia. Estuar Coast Shelf Sci 46:917-934 experimental evaluation for bicarbonate use in temperate Boss E, Zaneveld JRV (2003) The effect of bottom substrate on species. J Exp Mar Biol Ecol 265:203- 217 inherent optical properties: evidence of biogeocbemical Kelly MG (1980) Remote sensing of seagrass beds. In: Phillips processes. Llmnol Oceanogr 48:346-354 RC, McRoy CP (eds) Handbook of seagrass biology: an >- Brush MJ, N'u:on SW (2003) Biomass layering and metabolism ecosystem perspective. Garland STPM Press, New York, in mats of the macroalga Ulva lactuca L Estuaries 26: p69-85 916-926 >- Koch MS, Madden CJ (2001) Patterns of primary production Buonassissi C (2009) A regional comparison of particle size and nutrient availability in a Bahamas lagoon with hing distributions and the power-law approximation in oceanic ing mangroves. Mar Ecol Prog Ser 219:109-119 and estuarine surface waters by laser diffraction. MS >- Kutser T, Miller I, Jupp DLB (2006) Mapping coral reef ben thesis, University of Connecticut. Storrs, CT tbic substrates using hyperspectral space-borne images Burdige DG, Zimmerman RC (2002) Impact of sea grass and spectral libraries. Estuar Coast Shelf Sci 70:449-460 density on carbonate dissolution in Bahamian sediments. >- Lee ZP, Carder KL. Mobley CD, Steward RG, Patch JS (1999) Llmnol Oceanogr 47:1751-1763 Hyperspectral remote sensing for shallow waters: 2. Burdige D, Zimmerman R, Hu X (2008) Rates of carbonate Deriving bottom depths and water properties by optimiza dissolution in permeable sediments estimated from pore tion. Appl Optics 38:3831-3843 water profiles: the role of sea grasses. Llmnol Oceanogr Lee ZP, Hu C, Gray D, Casey Band others (2007) Properties of 53:549-565 coastal waters around the US: preliminary results using >- Dekker AG, Brando VE, Anstee JM (2005) Retrospective sea- MERIS data. Proc Envisiat Symp, Montreux, Switzerland, Dierssen et al.: Benthic productivity from space 15 23-27 April 2007 new lime mud budget. Geology 25:947-950 >- Lesser MP, Mobley CD (2007) Bathymeby, water optical prop Scoffin TP (1970) The trapping and binding of subtidal car erties, and benthic classification of coral reefs using hyper bonate sediments by marine vegetation in Bimini Lagoon. spectral remote sensing imagery. Coral Reefs 26:819-829 Bahamas. J Sediment Petrol 40:249- 273 >- Longhurst A. Sathyendranath S, Platt T, Caverhill C (1995) An Shinn EA. Steinen RP, Lldz BH, Swart PK (1989) Whitings. a estimate of global primary productivity in the ocean from sedimentologic dilemma. J Sediment Petrol 59:147-161 satellite radiometer data. J Plankton Res 17:1245- 1271 Short FT (1980) A simulation model of the seagrass produc Louchard EM, Reid RP, Stephens FC, Davis CO, Leathers RA. tion system. In: Phillips RC, McRoy CP (eds) Handbook Downes TV (2003) Optical remote sensing of benthic habi of seagrass biology: an ecosystem perspective. Garland tats and bathymeby in coastal environments at Lee Stock STPM Press, New York. p 277-296 ing Island, Bahamas: a comparative spectral classification >- Short FT, Wyllie-Echeverria S (1996) Natural and human approach. Limnol Oceanogr 48:511-521 induced disturbance of seagrasses. Environ Conserv 23: >-Macintyre HL, Geider RJ, Miller DC (1996) Microphyto 17-27 benthos: the ecological role of the 'secret garden' of un >- Smith SV (1981) Marine macrophytes as a global carbon sink. vegetated, shallow-water marine habitats. I. Distribution. Science 211:838-840 abundance and primary production. Estuaries 19:186-201 Stephens FC, Louchard EM, Reid RP, Mattione RA (2003) Maptech (2004) Region 9.2 Centnd Bahamas. Maptech, Effects of microalgal communities on reflectance spectra Amesbury, MA of carbonate sediments in subtidal optically shallow >- Maritorena S, Morel A, Gentili B (1994) Diffuse reflectance marine environments. Limnol Oceanogr 48:535-546 of oceanic shallow waters: inOuence of water depth and >- Stramski D, Boss E, Bogucki D, Voss KJ (2004) The role of sea bottom albedo. Limnol Oceanogr 39:1689-1703 water constituents in light backscattering in the ocean. Mateo MA. Cebrian J, Dunton K, Mutchler T (2006) Carbon Prog Oceanogr61:27-56 Dux in seagrasses. In: l.arkum AWD, Orth RJ, Duarte CM Stumpf RP, Holderied K. Robinson JA. Feldman G, Kuring N (eds) Seagrasses: biology, ecology, and conservation. (2003) Mapping water depths in clear water from space. Springer, Dordrecht, p 159-192 Proc 13th Biennial Coastal Zone Conf, NOAA Coastal >- Milliman JD (1993) Production and accumulation of calcium Services Center, Baltimore, MD, July 13-17, 2003, NOAA carbonate in the ocean: budget of a nonsteady state. Coastal Services Center, NOAA/CSC/20322-CD, Charles Global Biogeochem Cycles 7:927-957 ton, SC Mobley CD (1994) Light and water: radiative transfer in >- Twardowski MS, Boss E, Macdonald JB, Pegau WS, Barnard natural waters. Academic Press, San Diego AH, Zaneveld JRV (2001) A model for estimating bulk Mobley CD (1999) Estimation of the remote-sensing reflect refractive index from the optical backscattering ratio and ance from above-surface measurements. Appl Optics 38: the implications for understanding particle composition in 7442-7455 case I and case D waters. J Geophys Res 106:14129- 14142 >- Mobley CD, Sundman JK. Davis CO, Bowles JH and others >- Valiela I, McClelland J, Hauxwell J , Behr PJ, Hersh D, Fore (2005) Interpretation of hyperspectral remote-sensing man K (1997) Macroalgal blooms in shallow estuaries: imagery by spectrum matching and look-up tables. Appl controls and ecophysiological and ecosystem consequences. Opt 44:3576-3592 Llmnol Oceanogr 42:1105- 1118 Monahan E (1993) Occurrence and evolution of acoustically Vergara J, Perez-Llorens LP, Peralta G, Hernandez I (1997) relevant sub-surface bubble plumes and their associated, Seasonal variation of photosynthetic performance and remotely monitorable surface whitecaps. In: Kerman BR light attenuation in Ulva canopies from Palmones River (ed) Natural physical sources of underwater sound: sea Estuary. J Phycol 33:773-779 surface sound. Kluwer Academic Publishers, Dordrecbt. Winland HD, Matthews RK (1974) Origin and significance p 503-518 of grapestone, Bahama Islands. J Sediment Petrol 44: >- Otis DB, Carder KL, English DC, Ivey JE (2004) CDOM trans 921-927 port from the Bahamas Banks. Coral Reefs 23:152- 160 Zieman JC, Wetzel RG (1980) Productivity in seagrasses: >- Palacios SL. Zimmerman RC (2007) Response of eelgrass methods and rates. In: Phillips RC, McRoy CP (eds) Hand Zostera marina to c~ enrichment: possible impacts of book of seagrass biology: an ecosystem perspective. climate change and potential for remediation of coastal Garland STPM Press, New York. p 87-116 habitats. Mar Ecol Prog Ser 344:1-13 Zieman JC, Thayer GW, Robblee MB, Zieman RT (1979) Pro Petersen CGJ (1918) The sea bottom and its production of duction and export of seagrasses from a tropical bay. In: fish-food. Rep Dan Biol Stn 25:1-62 Livingston RJ (eds) Ecological processes in coastal and >- Phinn S, Roelfsema C, Dekker A. Brando V, Anstee J (2008) marine systems. Plenum Press, New York. p 21- 33 Mapping sea.grass species, cover and biomass in shallow Zimmerman R (2003) A biooptical model of irradiance distrib· waters: an assessment of satellite multi-spectral and ution and photosynthesis in seagrass canopies. Llmnol airborne hyper-spectral imaging systems in Moreton Bay Oceanogr 48:568-585 (Australia). Remote Sens Environ 112:3413- 3425 >- Zimmerman RC, Kohrs DG, Steller DL, Alberte RS (1997) >- Robbins LL, Tau Y, Evans CA (1997) Temporal and spatial dis Impacts of C~ enrichment on productivity and light tribution of whitings on the Great Bahama Bank and a requirements of eelgrass. Plant Physiol 115:599-607 Editorial responsibility: Rod.Dey Forster, Submitted: July 28, 2009; Accepted: May 12. 2010 Lowestott. UK Proofs received from autbor(s}: June 21, 2010